Image Processing for Face Authenticity via Dynamic Face-Region Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing authenticity determination methods based on facial feature points are vulnerable to spoofing attacks using photographs, and accurate determination results are difficult to obtain.
Innovation Solution
An image processing device that stores time-series images of an object changing facial orientation, extracts face regions, and determines authenticity based on the time-series changes in pixel values and aspect ratios of these images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If facial feature point movement is used for authenticity determination, then the determination process can be implemented, but spoofing attacks using photographs cannot be completely prevented
Solution Approach 1:
The patent transitions from analyzing only facial feature point movements (2D position changes) to incorporating aspect ratio changes of the entire face region (adding dimensional information). This multi-dimensional analysis includes both the movement trajectory of feature points and the dynamic changes in face region proportions, making photograph-based spoofing more detectable since static images cannot reproduce natural aspect ratio variations during facial movements.
Solution Approach 2:
The patent emphasizes dynamic analysis by capturing multiple frames during facial movements and analyzing temporal changes in both feature point positions and face region aspect ratios. This dynamic approach compares the temporal patterns of change against expected natural movement patterns, enabling detection of anomalies that occur with photographs being manipulated during the authentication process.
2Measurement precision
If only facial feature point movement is analyzed, then the processing complexity is low, but determination accuracy is insufficient
Solution Approach 1:
The patent segments the face region extraction process to obtain the bounding box of the entire face, then separately analyzes the aspect ratio of this face region in addition to feature point movements. This segmentation allows the system to extract and process multiple independent features (feature point coordinates and face region dimensions) without requiring a complete redesign of the authentication framework, thus improving accuracy while managing complexity.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A storage unit stores a plurality of time-series images that has captured an object when the object is instructed to change an orientation of the face. An action specifying unit extracts a face region from each of the plurality of images, obtains a change characteristic of pixel values of a plurality of pixels arranged in a predetermined direction in the face region, and specifies an action of the object on the basis of a time-series change in the change characteristics obtained from each of the plurality of images. A determination unit determines authenticity of the object on the basis of the action of the object.